Stability Dynamics of Main Qualitative Traits in Maize Cultivations across Diverse Environments regarding Soil Characteristics and Climate

Abstract

One of the main obstacles to finding cultivars with consistent performance across locations and years is the genotype × environment (GE) interaction effect. A new approach to stability analysis for qualitative characteristics in maize was conducted utilizing G × E interactions and further analysis via AMMI and GGE biplots. The study aimed to identify the type of trait inheritance through estimations of the stability index, to evaluate multiple locations and multiple genotypes to determine how different ecosystems and maize genotypes relate to one another, and, finally, to suggest the ideal climatic conditions and genotypes, carefully chosen for their stability. Fifteen F1 commercial maize hybrids comprised the genetic materials tested, along with 15 open-pollination lines created by 4-cycle Honeycomb assessment, at four different environments, Giannitsa, Florina, Trikala, and Kalambaka in Greece. The experiments were conducted in Randomized Complete Block Designs (RCB) with four replications. The tested characteristics were protein content (%), fat (%), ash (%), starch (%), crude fiber (%), moisture (%), seed length, seed thickness, and seed width. All genotypes showed statistically significant differences for all characteristics measured, especially for protein content and size of the kernel. G × E interaction was present only for moisture content and size of the kernel. Environments significantly affected fat, starch content, moisture content, and the kernel’s size (under a multiple G × E interaction). Protein, ash, and fiber content showed no G × E interaction. Further analysis via AMMI and GGE biplots was applied to explore the genotypic stability across all experimental environments for the traits that showed noteworthy G × E interaction. According to our results and approach, protein content is less qualitative than other characteristics like moisture and starch content. Correlations showed that negative selection for the last two characteristics, as well as for ash content, in combination with longer seeds, may lead indirectly to improved stability performance for protein content. Three environments, Giannitsa, Trikala and Kalambaka, exhibited higher stability index values for almost all characteristics measured. Therefore, those environments are perfect for ensuring the stability of the quality characteristics and could be recommended. The best maize hybrids were Mitic, 6818 and 6040, exhibiting high stability indices of quality characteristics and Kermes displaying stability for protein content. Therefore, those should be further tested in multiple environments to confirm the consistency of their high-stability performance.

1. Introduction

One of the most significant crops for many farmers in Greece is maize, soya, and peas, contributing to supporting livestock. According to the Food and Agriculture Organization (FAO) data, maize covered an area of 120,070 ha in 2021 in Greece, with a production of 1.322 million tones [1]. Qualitative traits are beneficial for defining the nutritional value of maize as an animal feed. Carbohydrates, proteins, fats, fiber, minerals, vitamins, and water are essential nutrients for animals to feed, grow, and breed [2]. Maize kernels usually contain a high quantity of starch, 6–12% protein and oil ranging from 3% to sometimes over 6% [3,4]. Studies of genotypes for high protein content are few compared to oil content [5]. Almost 49% of cultivated maize is used as raw material in the animal feed industry [6,7]. Protein and oil content are negatively related to each other and to starch content [8].

A genotype that performs well in one environment during one season or year may not perform well at a different time or another site within the same region because significant genotype-by-environment interactions (G × E) frequently occur under stress situations [9]. This is due to crossover performances from genotypes’ varying degrees of phenotypic expression under various environmental circumstances [10]. Differences in the sensitivity of the genotypes to the conditions in the target environment are another cause of genotype-environment interaction [10]. Therefore, developing better genotypes in various contexts depends heavily on genetic environment interactions (G × E).

Breeders need reliable data to know which genotypes to suggest, select, or reject. The multi-environmental trial (MET) study has extensively used the AMMI model and GGE biplots because they offer more precise estimates and make it simple to interpret the GEI through attractive graphical tools [11].

Different environments affect performance significantly. Thus, maize breeders have to improve new genotypes in terms of performance stability for vegetative and yielding characteristics [12,13]. The strength of maize performance is crucial for proper hybrid selection in various cultivation systems and environments to ensure better performance of genotypes [14]. Stability is referred to as the ability of a genotype to perform satisfactorily under different environments and cope with various biotic and abiotic stresses [15]. Fasoula’s [15] and Fasoula [16] used the Coefficient of Variation as a stability criterion in many cases to describe genotype adaptation under different environments. Maize stability under varying densities was proposed by Ipsilandis and Vafias [17]. Also, Vafias et al. [14] depicted that the most severe stress for maize genotypes was low-input farming that suppressed genotype expression. Greveniotis et al. [18,19] published an implemented method for stability estimations, in maize and other crops, based on Fasoula’s [20,21] approach for selecting stable genotypes that exhibit high performance in various environments. Also, Greveniotis et al. [22,23,24,25] went further in exploring different traits of peas (Pisum sativum L.) and common vetch (Vicia sativa L.), utilizing AMMI and GGE biplot analyses as tools. Fasoula [16,21] used selection criteria for stability according to the squared form of the reversed coefficient of variation, i.e., the ratio ${(\overline{x}/s)}^2$. Kang [26] reported that breeders must aim at improving high-yielding performance and stability of performance simultaneously. Other authors [27] have assessed different maize genotypes in Greek environments based on grain quality traits and applied AMMI analysis along with Principal Component Analysis. However, to our knowledge, it is the first time that the stability of maize genotypes regarding seed quality and chemical composition parameters was assessed in several diverse environments in Greece, applying new assessment methodology, i.e., Fasoula’s stability index and utilizing the AMMI and GGE biplot analyses as further tools.

The purposes of this study were: (a) to use Fasoula’s stability criterion for the simultaneous selection of best and stable genotypes based on seed quality and chemical composition parameters, (b) through estimations of the stability index to determine the type of trait inheritance, (c) to evaluate multiple locations and multiple genotypes to identify how different environments and maize genotypes are related, and, finally, (d) to suggest the ideal conditions and genotypes, carefully chosen for their stability.

2. Materials and Methods

2.1. Crop Establishment and Experimental Procedures

Four locales in the northern and central parts of the country (Figure 1) were chosen for the installation of two-year experimentation. The selection of locales was based on the availability of farming land and the diversity of geographical position, elevation, soil texture and climate, as displayed in Table 1 and Figure 2, respectively.

Figure 1. Map showing the location of four environments.

Table 1. Geographical location, elevation, soil chemical analysis and crop information for the experimental plots.

Environments	Longitude	Latitude	Elevation (m)	Sand	Silt	Clay	Soil Texture	pH	Organic Matter	Planting Date	Harvesting Date
Giannitsa	22°39′ E	40°77′ N	10	8.9%	37.4%	53.7%	clay (C)	8.18	3.50%	middle April 2011 and 2012	late September 2011 and 2012
Florina	21°22′ E	40°46′ N	705	61.2%	27.6%	11.2%	sandy loam (SL)	6.25	1.29%	early May 2011, middle May 2012	middle December 2011, early November 2012
Trikala	21°64′ E	39°55′ N	120	49.0%	19.0%	32.0%	Sandy clay loam (SCL)	8.0	2.40%	early April 2011 and 2012	Early October 2011 and 2012
Kalambaka	21°65′ E	39°64′ N	190	1.6%	49.1%	49.3%	silty clay (SiC)	8.18	2.14%	early April 2011 and 2012	Early October 2011, late September 2012

Figure 2. Monthly mean temperature (in °C) and precipitation (in mm) for growing seasons 2011 and 2012, at the areas of experimentation.

The genetic materials were consisted of 15 F1 commercial maize hybrids, i.e., G1 was PR31Y43, G2 was FACTOR, G3 was COSTANZA, G4 was ARMA, G5 was PR31A34, G6 was ELEONORA, G7 was FAMOSO, G8 was DKC6818, G9 was MITIC, G10 was DKC6040, G11 was KERMESS, G12 was PR31G98, G13 was LG3535, G14 was PR33A46 and G15 was PR31P41 and 15 open-pollination lines that were created between 2007 and 2010 after a 4-cycle Honeycomb assessment with a selection intensity close to 1%., i.e., G16 was 1T, 2T was G17, G18 was 3T, G19 was 4T, 5T was G20, G21 was 1F, G22 was 2F, G23 was 3F, G24 was 4F, G25 was 5F, G26 was 6F, G27 was 7F, G28 was 8F, G29 was 9F and G30 was 10F. The F2 (C0) generation of the F1 commercial maize hybrid Costanza has been utilized for open-pollinated lines. The pedigree selection method was used throughout the experimentation on the half-sib offspring regardless of open pollination.

The final trials included four replications and were conducted using the Randomized Complete Block Design (RCB). [28]. The plots had a row-to-row spacing of 75 cm and comprised two single rows. Plant-to-plant spacing was 18 cm, with 25 plants in each row. To reach the necessary plant densities, hills were over-seeded at a double rate and then manually thinned when the maize crop reached the two-leaf stage.

To guarantee a good plant stand, all relevant agronomic and cultural procedures were timely followed. After seeds were sown, the same amounts of nitrogen and phosphorus fertilizer (element level) were administered in all four locations and experimental plots: 150 and 75 kg/ha. Additional nitrogen fertilizer (135 kg/ha) was then applied when the plants reached a height of 50 cm (boot stage). Herbicides applied post-emergence ensured weed control. To prevent water stress at any growth stage, irrigation was carried out regularly (often every ten days). Each plot was picked by hand to ensure excellent accuracy.

2.2. Measurements

After harvest, the seeds from each genotype per plot at a given location were bulked. 100 g maize seed samples were collected and analyzed for quality and chemical composition parameters. All characteristics were measured in the Laboratory of Animal Technology at the University of Thessaly.

For chemical analysis, approximately 50 g of dry maize seeds (13% moisture) were cleaned, and ground using a laboratory mill and were stored at 4 °C until analysis.

Approved methods of the American Association of Cereals Chemists (AACC) were applied for the determination of fat (%) (Soxhlet extraction with petroleum ether based on AACC Method 30-25.01), ash (%) (AACC Method 08-01), crude protein (CP) (%) (determination of total nitrogen by Kjeldahl method according to modified AACC Method 46-12 and expression of protein content as % CP = total N × 6.25), moisture (%) (AACC Method 44-15A) [29]. More specifically, 3 g of ground homogenized seed sample previously dried for fat determination was extracted by Soxhlet extraction using petroleum ether as the solvent for 4 h. The solvent evaporates by a rotary evaporator, and the residue is dried at 100 degrees to a consistent weight. The residue is expressed as % crude fat or ether extract [29]. For ash determination, 3 g of ground homogenized seed samples were weighed in a porcelain crucible previously ignited, cooled in a desiccator, and weighed after reaching room temperature. Next, the sample is ignited in a furnace at 550 °C until constant weight, cooled in a desiccator and weighed when it reaches room temperature [29].

Official Methods by the Association of Official Analytical Chemists (AOAC) were applied for the determination of crude fiber (%) (AOAC 991.43) and total starch (%) (AOAC method 996.11 [30]. In addition, mean values of seed length, seed thickness, and seed width (in mm) were obtained by measuring the three dimensions of 20 randomly selected seeds with a Vernier Calliper.

2.3. Data Analysis

To determine if there are significant differences for all variables examined in this study, data were largely evaluated by ANOVA over locations and years. The experiments were conducted in Randomized Complete Block Designs (RCB) with four replications. The environment was designated as the product of each year and location to make the ANOVA table more illuminating. By doing this, we reduce the number of interactions in the ANOVA table and maintain the G × E (Genotype × Environment) interaction, which is essential for advancing the stability study.

For stability estimations, the presented ratio ${(\overline{x}/s)}^2$, where $\overline{x}$ and s represent the entry mean yield and standard deviation, respectively, is used, namely the stability index [16,21]. Pearson coefficient based on Steel et al. [28] was employed to search for trait correlations, and SPSS ver. 25 software for statistics was used to explore all data for statistical significance at p < 0.05. The PB Tools v.1.4—free version (International Rice Research Institute, Laguna, Philippines) was utilized for computing the AMMI and GGE biplot analysis for interactions.

3. Results

In Table 2, ANOVA main results are presented. All genotypes showed statistically significant differences for all characteristics measured, especially for protein content and size of the kernel. G × E interaction was present only for moisture content and size of the kernel. Environments significantly affected fat, starch content, moisture content, and the kernel’s size (which was under a multiple G × E interaction). Protein content, ash and fibers showed no G × E interaction.

Table 2. Mean squares for the tested seed quality and chemical composition parameters from the analysis of variance across environments.

Source of Variation	Protein Content (%)	Fat Content (%)	Ash Content (%)	Starch Content (%)	Crude Fiber Content (%)	Moisture Content (%)	Seed Length (mm)	Seed Thickness (mm)	Seed Width (mm)
Environments (E)	0.122 ns	0.016 *	0.001 ns	1.197 **	0.021 ns	0.212 **	0.238 **	0.137 **	0.069 *
REPS/Environments	0.055 ns	0.006 ns	0.001 ns	0.534 ns	0.016 ns	0.003 ns	0.071ns	0.055 ns	0.048 ns
Genotypes (G)	11.910 **	0.579 **	0.113 **	4.615 **	0.112 **	0.105 **	20.158 **	2.002 **	1.419 **
Genotypes × Environments (G × E)	0.038 ns	0.008 ns	0.001 ns	0.470 ns	0.012 ns	0.204 **	0.109 **	0.051 *	0.051 *
Error	0.075	0.007	0.001	0.420	0.017	0.003	0.064	0.044	0.039

Probability values: * p ≤ 0.05; ** p ≤ 0.01; ns = not significant.

Table 3, Table 4, Table 5 and Table 6 present the trait stability index ${(\overline{x}/s)}^2$ values for protein content (%), fat content (%), ash content (%), starch content (%), crude fiber content (%), moisture content (%), seed length, seed thickness, seed width, in all four different environments. Moisture content and starch content showed the greatest stability index values. Especially moisture content showed extreme values close to or over 100,000 (i.e., 126,070 in Florina, 93,347 in Giannitsa). In contrast, crude fiber content and seed width showed values below 1000. Finally, protein content and general seed size showed low values, just above or near 1000.

Table 3. Trait stability index ${(\overline{x}/s)}^2$ for seed quality and chemical composition parameters in Giannitsa.

GENOTYPES	Protein Content (%)	Fat Content (%)	Ash Content (%)	Starch Content (%)	Crude Fiber Content (%)	Moisture Content (%)	Seed Length (mm)	Seed Thickness (mm)	Seed Width (mm)
31Y43	1176	1589	3592	9110	228	21,040	770	2828	530
FACTOR	1124	2404	2239	5781	426	5771	1131	2610	298
COSTANZA	859	874	6623	7991	488	38,183	909	2993	1037
ARMA	858	1093	2342	7428	170	22,708	1765	4138	1087
31A34	984	1051	2154	31,219	256	4167	1807	4410	669
ELEONORA	1431	2298	1533	6330	337	66,523	1590	1445	446
FAMOSO	1081	2256	2554	9751	276	1915	1724	1367	1776
6818	949	1894	2289	27,505	277	1131	1917	3981	1165
MITIC	916	1142	1957	11,082	409	5428	2295	1180	2983
6040	1173	1310	3252	12,045	481	1103	1297	3710	661
KERMESS	1062	2003	2984	7023	147	40,125	817	2477	1015
31G98	846	925	2945	11,546	265	3424	920	774	350
LG3535	1228	962	2712	9703	290	4315	3803	2574	592
33A46	743	2511	2561	11,140	326	7435	2324	879	335
31P41	1273	1678	2062	11,112	417	1409	2641	14,372	658
1T	754	1509	2004	20,134	326	4172	2832	2212	470
2T	1017	1346	3104	6308	336	1357	5618	1939	1678
3Τ	1022	835	2360	8201	382	35,384	3925	1836	973
4Τ	1138	1032	3229	5619	221	93,347	3889	1704	494
5Τ	1115	1324	2683	5942	363	4423	2518	4305	646
1F	966	790	2350	7739	290	1723	1895	4139	630
2F	931	1172	2851	23,876	434	3154	1709	1504	572
3F	1233	1633	3134	11,794	236	28,987	1471	2515	994
4F	1040	1673	846	24,707	219	10,148	1215	3245	521
5F	788	1709	2269	17,964	430	1488	2940	1034	566
6F	914	1218	2638	6032	260	1541	2360	1745	365
7F	885	421	3247	4406	249	43,753	984	960	341
8F	855	1058	2504	5924	430	2514	1249	1781	878
9F	945	1476	2225	8012	287	13,359	638	2227	536
10F	976	1515	3038	29,253	303	1536	1902	3633	728

Table 4. Trait stability index ${(\overline{x}/s)}^2$ for seed quality and chemical composition parameters in Florina.

GENOTYPES	Protein (%)	Fat (%)	Ash (%)	Starch (%)	Crude Fiber (%)	Moisture (%)	Seed Length (mm)	Seed Thickness (mm)	Seed Width (mm)
31Y43	853	1607	1709	7115	199	13,063	3442	597	430
FACTOR	750	2173	9280	5087	206	4825	610	2656	449
COSTANZA	1023	1430	1626	14,170	256	126,070	576	1872	313
ARMA	1443	1078	3321	15,815	559	24,051	1320	879	387
31A34	1155	2026	6946	7199	511	6112	3088	2229	545
ELEONORA	832	1192	2223	7191	329	33,084	1873	992	394
FAMOSO	906	856	2767	12,740	613	2051	1128	2008	493
6818	1111	1376	3639	29,684	314	1611	3077	1624	3389
MITIC	939	1788	2175	32,629	527	12,197	1634	787	264
6040	1223	3155	2087	20,412	415	1292	2723	4029	1218
KERMESS	937	1469	2889	10,016	392	59,319	1510	2005	1060
31G98	1392	3010	2554	6063	359	2527	1417	3140	767
LG3535	1015	1816	2752	7254	471	5254	694	1706	575
33A46	1245	1284	3812	7888	347	9634	2575	1006	660
31P41	905	1727	2678	7105	312	1468	2375	1319	808
1T	1255	1670	5209	4720	299	4043	1106	1662	371
2T	1056	1089	2867	6393	237	1327	2529	790	358
3Τ	980	1648	2163	6467	326	17,108	2378	2620	668
4Τ	867	2449	2605	8814	487	65,561	1919	943	375
5Τ	1042	1947	3620	9327	346	3946	1163	1685	352
1F	1062	3774	5189	7582	398	1808	1693	14,877	326
2F	1169	1541	2308	7425	346	3014	2700	850	498
3F	1002	1527	2245	4800	465	60,878	2636	3231	499
4F	958	962	1794	6683	573	6925	1968	2650	510
5F	1201	1005	3993	8228	294	1754	1512	2929	811
6F	1410	1202	2470	15,470	420	1447	2686	2430	1014
7F	1159	570	2916	8300	582	19,503	1273	2076	967
8F	1292	2914	2905	10,091	327	2143	883	2263	849
9F	1085	2027	2471	5557	297	16,786	4450	1773	1744
10F	1200	2181	3236	13,005	423	1789	2319	1143	379

Table 5. Trait stability index ${(\overline{x}/s)}^2$ for seed quality and chemical composition parameters in Trikala.

GENOTYPES	Protein (%)	Fat (%)	Ash (%)	Starch (%)	Crude Fiber (%)	Moisture (%)	Seed Length (mm)	Seed Thickness (mm)	Seed Width (mm)
31Y43	1009	1261	883	8747	406	7352	1768	3101	238
FACTOR	502	2191	2519	3499	287	26,327	866	2146	191
COSTANZA	660	1201	3913	14,386	558	2665	2395	3956	616
ARMA	965	1499	2342	9812	247	21,293	2168	1990	271
31A34	842	2481	10,199	14,812	288	25,310	3689	975	379
ELEONORA	935	1429	1611	10,697	423	3727	1306	3768	1051
FAMOSO	787	1644	3601	31,467	268	12,988	1128	1506	452
6818	799	796	3044	13,060	222	2997	1721	1312	973
MITIC	883	2136	1910	15,236	462	25,540	1890	2571	2452
6040	1369	2566	3191	11,663	339	25,965	1114	3122	637
KERMESS	1353	1218	2446	9176	294	42,910	2802	1701	380
31G98	896	1870	3835	7531	805	1584	7007	926	345
LG3535	761	3372	2712	8299	522	4688	1085	1738	429
33A46	774	1434	2893	18,705	336	7183	1437	1152	413
31P41	1178	1048	2839	6781	780	22,095	2329	824	330
1T	1545	1798	2248	14,429	315	7279	2312	1000	432
2T	1332	1941	2259	11,303	334	13,322	1786	736	1673
3Τ	830	1044	3542	10,062	260	3244	3744	1260	770
4Τ	1086	1309	2360	7771	187	4443	2794	3636	1387
5Τ	1252	1006	2804	7156	168	2486	6734	1240	311
1F	1086	1861	3176	6012	311	11,634	4261	2244	550
2F	1136	1844	2499	9319	320	5098	2022	3479	1050
3F	989	1478	2922	13,209	355	1434	899	2472	386
4F	927	1301	2460	6845	362	5293	1441	1501	889
5F	1021	1245	3838	7837	436	2852	2647	1132	740
6F	1155	1584	3680	16,028	482	28,191	1676	1290	392
7F	960	416	3010	3899	330	1199	1216	1345	371
8F	1296	1462	2330	10,464	536	5391	858	918	895
9F	955	1984	4397	7052	244	11,259	1416	1138	388
10F	1554	1412	3140	9867	170	15,922	1797	3854	264

Table 6. Trait stability index ${(\overline{x}/s)}^2$ for seed quality and chemical composition parameters in Kalambaka.

GENOTYPES	Protein (%)	Fat (%)	Ash (%)	Starch (%)	Crude Fiber (%)	Moisture Content (%)	Seed Length (mm)	Seed Thickness (mm)	Seed Width (mm)
31Y43	668	2054	1687	6159	107	5206	1340	2178	309
FACTOR	1349	1735	3064	10,733	239	34,483	1450	630	731
COSTANZA	1018	2840	2157	8782	265	3705	1007	4240	192
ARMA	1030	2210	3823	12,293	361	7744	939	1253	602
31A34	1127	1784	1627	8753	597	100,640	4511	1890	226
ELEONORA	1028	1255	2774	7575	348	4034	5382	2122	899
FAMOSO	985	2083	2405	16,974	317	9429	757	1698	381
6818	968	1646	2678	23,748	2906	4354	1907	935	777
MITIC	1042	1643	5023	29,778	427	14,756	2616	2377	847
6040	1041	1660	3625	31,207	597	37,199	1408	2139	490
KERMESS	3003	2115	2840	8543	370	16,848	2587	2241	894
31G98	991	1884	2481	16,875	221	1427	1041	2067	461
LG3535	1367	1196	2848	11,483	255	3246	1713	1426	543
33A46	1725	1493	2709	13,853	476	7960	1936	1960	2772
31P41	1112	1110	2119	9457	234	24,804	3370	1492	702
1T	1280	1706	4040	17,896	348	7518	2189	1512	949
2T	1338	1532	3612	7988	366	23,123	2544	872	333
3Τ	1113	2565	2101	5575	349	2727	964	2571	348
4Τ	1122	1820	4471	8527	618	4147	1175	3297	871
5Τ	986	1813	2836	12,844	617	2074	2323	2045	878
1F	830	1822	2886	8626	478	18,492	1150	1118	882
2F	946	1644	3540	8917	422	5658	1072	2750	598
3F	1176	1672	2214	5008	346	1350	6812	782	477
4F	1312	1849	3616	11,926	364	3147	5802	2173	315
5F	1087	2084	1638	5134	267	2768	816	1337	702
6F	1009	2021	2662	3804	270	29,753	1961	1686	329
7F	1263	463	3460	3398	495	1220	808	8926	1125
8F	999	1290	2758	11,715	296	4925	1002	2261	432
9F	1283	1173	2113	8861	393	14,385	2035	2602	476
10F	935	1911	3460	13,343	588	15,926	1984	979	394

Table 7 and Table 8 show indices across environments and genotypes, respectively. Giannitsa, Trikala and Kalambaka exhibited higher indices for almost all characteristics measured. From Table 8, some genotypes showed high indexes, especially for moisture and starch content (such as Mitic, 6818 and 6040). Also, some genotypes showed high stability indices for many other characteristics in specific environments (i.e., Costanza in Giannitsa). Again, crude fiber content and seed size showed low values for every genotype and environment. Hybrid Mitic showed higher values for almost all characteristics. The hybrid Factor showed low values for many characteristics, while lines of the F-series also demonstrated satisfactory results, indicating a stability potential.

Table 7. Trait stability index ${(\overline{x}/s)}^2$ across environments for seed quality and chemical composition parameters.

Environments	Protein Content (%)	Fat Content (%)	Ash Content (%)	Starch Content (%)	Crude Fiber Content (%)	Moisture Content (%)	Seed Length (mm)	Seed Thickness (mm)	Seed Width (mm)
Giannitsa	169	391	385	7306	266	2796	153	696	240
Florina	172	453	388	6127	305	2928	143	621	274
Trikala	168	430	366	7255	287	3505	166	744	227
Kalambaka	167	436	388	6672	265	3495	178	709	262
Real Mean	169	428	382	6840	281	3181	160	693	251

Table 8. Trait stability index ${(\overline{x}/s)}^2$ across genotypes for seed quality and chemical composition parameters.

GENOTYPES	Protein Content (%)	Fat Content (%)	Ash Content (%)	Starch Content (%)	Crude Fiber Content (%)	Moisture Content (%)	Seed Length (mm)	Seed Thickness (mm)	Seed Width (mm)
31Y43	877	1634	1649	7671	192	9258	1181	1256	283
FACTOR	879	1719	3212	5506	257	8016	934	1243	328
COSTANZA	937	1329	2815	11,479	333	6317	582	2704	358
ARMA	1054	1460	3029	10,488	287	13,974	1465	1610	399
31A34	1045	1491	3218	10,720	335	8142	2979	1829	407
ELEONORA	1086	1493	2103	7737	369	7570	1775	1779	633
FAMOSO	957	1614	3036	13,545	358	3648	1112	1698	558
6818	1001	1391	2999	18,116	347	2107	2120	1291	1032
MITIC	1038	1711	2503	19,583	448	10,948	1848	1274	582
6040	1272	1575	3206	17,328	460	2389	1249	2257	698
KERMESS	1345	1517	3014	9315	277	3217	1671	2226	740
31G98	971	1716	2958	9229	348	2134	1424	1332	418
LG3535	1125	1617	2844	9095	384	3338	1340	1908	518
33A46	1077	1657	2993	12,258	382	3786	2112	1156	559
31P41	1194	1455	2619	8728	343	2973	2043	1415	524
1T	1220	1445	3096	10,135	339	3388	1898	1596	504
2T	1269	1530	3151	8280	340	2612	2348	900	610
3Τ	1075	1402	2639	6740	331	2429	2160	1153	544
4Τ	1122	1549	3238	5696	289	2331	1959	1981	675
5Τ	1159	1224	3234	7732	316	2508	1678	1984	484
1F	1061	1654	3384	7878	353	3414	1936	2046	538
2F	1097	1649	2890	10,307	399	3496	1782	1544	613
3F	1170	1618	2784	7520	339	2848	1686	1718	531
4F	1135	1380	1565	9529	355	2647	1705	2399	524
5F	1090	1405	2776	8862	372	1953	1600	1297	526
6F	1169	1572	2950	7458	349	3125	2114	1626	462
7F	1128	490	3382	3756	347	2459	977	1733	556
8F	1189	1531	2787	9824	408	2478	995	1582	748
9F	1107	1744	2825	7129	312	2168	1381	1912	531
10F	1128	1641	3252	14,061	317	3243	2181	1621	353

In Table 9, numerous statistically significant correlations were found, particularly between characteristics of the kernel (seed) size (up to 0.633). Ash content was strongly and positively correlated to seed size (0.452 to 0.592). The same was found for fat content. Moisture content (%) was negatively correlated to essential characteristics such as protein, fat, fiber, and ash content. Only starch content was positively correlated to moisture content. Protein content generally was negatively correlated to all other characteristics, except for seed length, where the correlation was positive. Generally, larger seeds favor fat, ash, and fiber content but not starch, protein, and moisture.

Table 9. Correlations between all seed quality and chemical composition parameters.

	Protein (%)	Fat (%)	Ash (%)	Starch (%)	Fiber (%)	Moisture Content (%)	Seed Length (mm)	Seed Thickness (mm)
Fat (%)	0.046
Ash (%)	−0.134 **	0.283 **
Starch (%)	−0.068 *	0.065 *	−0.206 **
Fiber (%)	0.037	0.203 **	0.195 **	−0.024
Moisture content (%)	−0.069 *	−0.076 *	−0.034	0.077 *	−0.082 *
Seed length	0.104 **	0.331 **	0.592 **	−0.233 **	0.226 **	−0.077 *
Seed thickness	−0.144 **	0.200 **	0.523 **	−0.218 **	0.169 **	−0.026	0.633 **
Seed width	−0.004	0.240 **	0.452 **	−0.209 **	0.103 **	−0.016	0.574 **	0.633 **

* Correlations are significant at the 0.05 level (2-tailed), ** Correlation is significant at the 0.01 level (2-tailed).

The AMMI Tool for Multi-Environment Evaluations

The AMMI analysis is a statistical tool that is helpful for multi-environment experimental analysis in a manner that is easily understandable and can depict the complexity of GEI. For this reason, the data of GEI means are arranged in a two-dimension table. The (LS) least squares, estimated by the analysis, are used for generating a two-dimensional ANOVA model for the additive based on main effects and estimating a value decomposition on the residuals of interaction [31].

The statistical AMMI software tool depicts the adaptation figure and the AMMI1 biplot where the two axes are. The x-axis depicts the values of the factor analyzed (axis of the factor), and the y-axis is the one where the PC1 values are depicted. When the values of PC1 on the biplot are placed close to the axis of the factor (x-axis), it means that the factor analyzed expresses stability for all experimental environments. Therefore, the desirable genotypes on the AMMI1 biplot are those representing high values on the x-axis (the axis of factor performance) (x-axis, right position) combined with a low estimation of the PC1 y-axis (near zero).

The GGE biplot analysis combines the main effect of genotype (G) and the interaction of genotype by environment (GE). This combination is the core of stability analysis. A GGE biplot displays the genotype by the environment in the two-dimension data matrix. The methodology of the GGE biplot derives from the analysis by graphical tools of multi-environment trials of genotypes (MET).

In the GGE biplot for the characterization of the environments, the stable one was the one depicted near the ideal and average environment dots. Regarding the genotypes and their stability, the desirable genotypes were those depicted in the area of the ideal genotype.

The analysis using the AMMI1 and G×E biplot creates biplots based on the genotypic performance among environments. The created biplots are a simple tool that can easily categorize each genotype for stability and performance.

Figure 3a–d presents the analysis for the stability of moisture content (%) utilizing both AMMI and GGE biplot.

Figure 3. Analysis for stability for the trait of moisture content (%) based on (a) adaptation map; (b) AMMI1 biplot; (c) GGE biplot depicting the environmental stability; (d) GGE biplot for genotypes depicting the genotypic stability. The desirable genotypes are those placed near the concentric region of the ideal genotype. The E signs represent the environments combined with the years used as follows E1 was Giannitsa, E2 was Florina, E3 was Trikala, and E4 was Kalambaka for 2011; E5 was Giannitsa, E6 was Florina, E7 was Trikala and E8 was Kalambaka for the year 2012. G signs represent the genotypes used as follows G1 was PR31Y43, G2 was FACTOR, G3 was COSTANZA, G4 was ARMA, G5 was PR31A34, G6 was ELEONORA, G7 was FAMOSO, G8 was DKC6818, G9 was MITIC, G10 was DKC6040, G11 was KERMESS, G12 was PR31G98, G13 was LG3535, G14 was PR33A46, G15 was PR31P41, G16 was 1T, 2T was G17, G18 was 3T, G19 was 4T, 5T was G20, G21 was 1F, G22 was 2F, G23 was 3F, G24 was 4F, G25 was 5F, G26 was 6F, G27 was 7F, G28 was 8F, G29 was 9F and G30 was 10F.

The analysis for stability using AMMI and GGE biplot for seed length is shown in Figure 4a–d.

Figure 4. Analysis of stability for the seed length characteristic based on: (a) adaptation map; (b) AMMI1 biplot; (c) GGE biplot depicting the environmental stability; (d) GGE biplot for genotypes depicting the genotypic stability.

Figure 5a–d shows the analysis for the stability of seed thickness using AMMI and GGE biplot.

Figure 5. Analysis of stability for the seed thickness characteristic based on: (a) adaptation map; (b) AMMI1 biplot; (c) GGE biplot depicting the environmental stability; (d) GGE biplot for genotypes depicting the genotypic stability.

The analysis for stability using AMMI and GGE biplot for the trait of seed width is presented in Figure 6a–d.

Figure 6. Analysis of stability for the seed width trait based on: (a) AMMI adaptation map; (b) AMMI1 biplot; (c) GGE biplot for the environment; (d) GGE biplot for genotypes depicting the stability of the genotypes.

Based on the data of the adaptation map of AMMI analysis, the genotypes with the highest values on the performance axis of the trait, showing a more or less parallel diagram line to the axis of PC1, were desirable, as this was an indication of stability among environments.

Based on the biplot of AMMI1, the stable genotypes expressed high values on the trait performance axis (right position on the x-axis) and near the lowest values on the PC1 axis.

Regarding the environments and the biplot of the GGE, the stable environment was placed near the ideal and average environment and the dot of the ideal environment.

For the biplot of GGE regarding the genotypes, the data showed that the productive and stable (desirable ones) were those placed in the area of the ideal genotype.

4. Discussion

Farmers, especially when they possess livestock, want high and stable yields for their cultivations, with excellent and stable quality that may ensure the high capabilities of their animals [32].

Our research showed that moisture and starch content had the greatest stability index values. These two characteristics may easily ensure stability because they exhibit qualitative gene behavior and inheritance [18,19]. However, although this may be a fact, moisture content and seed size showed strong G × E interaction, revealing their vulnerability against environmental conditions. Also, fat and starch content presented different behavior in different environments. Conversely, protein, ash and fiber content showed no G × E interaction.

In maize, the variation in protein content was found to be lower than the variation in oil content. The opposite was reported by Aliu et al. [33], who reported that the protein and oil contents ranged between 11.02 to 13.02% and 2.56 to 5.57%, respectively. Has et al. [34] observed variation from 11.2 to 15.6%, and Prasanna et al. [35] presented similar results (from 8.9 to 10.2%). The starch content of maize kernels depends on genotype [33]. Similar results for starch content were reported by Has et al. [34]. Also, ash content (AC) ranged from 1.28 to 1.45% [33]. The feed industry must have high protein content in animal feed, especially for poultry and swine [36,37,38]. Therefore, a balanced maize feed must contain high protein and have the proper proportions of oil, starch, fat, water, etc., to be accepted as a complete feed. In our study, we tried to depict the most stable characteristics of maize kernel and the best environments and genotypes to achieve the best result.

In Giannitsa, Trikala, and Kalambaka, nearly all of the seed characteristics that were evaluated had higher indices. That suggests that particular environments are excellent for ensuring the stability of high-quality features. Mitic, 6818, and 6040 were the best maize hybrids, all with high stability indices. Additionally, several genotypes displayed particular adaptation in certain conditions as seen by high stability indices for various characteristics (e.g., hybrid Costanza in Giannitsa). Hybrid Kermess showed high stability for protein content.

Essential properties like protein and fat content and fiber and ash level were inversely linked with moisture content. Only starch content positively correlated with moisture content, maybe due to soluble sugars’ behavior. Except for seed length, protein content often had a negative association with all other variables.

Generally, larger seeds favor fat, ash, and fiber content but not starch, protein, and moisture. Scrob et al. [39] reported positive correlations found between ash content and protein content, test weight, and protein content and 1000-kernel weight among tested corn hybrids. In contrast, a negative correlation was found between starch and protein and starch and ash content. Saleem et al. [40] reported that oil and protein content are negatively correlated. They also reported that an increase in sugar contents might decrease the starch level of grain, which consequently reduces the grain yield. A negative and significant correlation was found between ash contents and grain yield at the genotypic and phenotypic levels.

AMMI and GGE biplot contributed to defining the best combination of the genotypes used and the environments/type of cultivation for the main seed characteristics studied in our research.

4.1. Moisture Content (%)

Figures of the AMMI1 biplot (Figure 3b) and adaptation map (Figure 3a) were created based on the trait of moisture content (%) using AMMI analysis. The AMMI figures explained a percentage of the total variability (45.1%), which is enough for conclusions to be drawn. Both AMMI1 and adaptation map figures show that genotypes G4 (ARMA) and G23 (3F) were the most stable, followed by G24 (4F), G29 (9F) and G18 (3T). The moisture % performance of each of the aforementioned genotypes was quite constant, which is desirable for the quality of the seeds. The trait’s range was 12.35 to 12.60%, which demonstrates the most desirable trait’s environmental stability. The analysis of GGE explained a portion of the whole variability of 74.3% (PC1:42.1%, PC2: 32.2%), which is estimated, very high. The Environment view of the GGE biplot (Figure 3c) revealed that no environment was similar. In contrast, E6 (Florina) and E7 (Trikala) environments were depicted in the area of the average environment. The genotype view of the GGE biplot (Figure 3d) shows that genotypes G28 (8F), G27 (7F), G23 (3F), G8 (DKC6818), G6 (ELEONORA) depicted identical with the ideal genotype and characterized by high stability, although exhibiting lower performance for this trait.

4.2. Seed Length

Regarding the trait of seed length, the AMMI1 biplot and Adaptation map analysis (Figure 4a,b) expressed a PC1:32.9% of the whole variability. In the two figures, it was apparent that the genotypes G1 (PR31Y43), G2 (FACTOR), G7 (FAMOSO) and G3 (COSTANZA) were more or less stable. The analysis of the GGE biplot explained a portion of 97.8% (PC1:96.6%, PC2: 1.2%) of the total variability. The Environment view GGE biplot (Figure 4c) revealed that all the seed length trait experimental environments were very close and highly stable. Every environment fits into the in-centric area of an average and ideal environment. The Genotypes view of the GGE biplot (Figure 4d) shows that G3 (COSTANZA), G2 (FACTOR), G1 (PR31Y43) and G7 (FAMOSO) were rather stable genotypes and were all positioned within the ideal genotype’s concentric circles.

4.3. Seed Thickness

The analysis of AMMI for the seed thickness trait explained for PC1 a portion (38.7%) of the whole variability. Both the AMMI1 biplot (Figure 5b) and Adaptation map (Figure 5a) and shown that the relatively stable genotypes were the G1 (PR31Y43), G2 (FACTOR), G3 (COSTANZA) and G5 (PR31A34). The analysis of the GGE biplot explained 91.6% (PC1:85.8%, PC2: 5.8%) of the whole variability. The environmental view of the GGE biplot revealed that all environments for the seed thickness trait were extremely close and very stable. All environments were positioned into the concentric area of average and ideal environments.

The biplot of GGE for genotype view shows that genotypes G1 (PR31Y43), G2 (FACTOR), G3 (COSTANZA) and G13 (LG3535) were the most desirable, as they were all placed within the concentric area of ideal genotype.

4.4. Seed Width

The analysis of AMMI, as presented from the AMMI1 biplot and Adaptation map figures, explained a PC1: 29.2% of the whole variability. AMMI1 biplot (Figure 6b) and Adaptation map (Figure 6a) showed that the genotypes G2 (FACTOR), G3 (COSTANZA), G5 (PR31A34) and G4 (ARMA) were the most stable. The analysis of the GGE biplot explained a portion of 86.5% (PC1:80.7%, PC2: 5.8%) of the whole variability. The biplot of GGE for environment view revealed that every environment was positioned near the average and the ideal environment. Therefore, all environments were quite similar and stable. The genotype view of the GGE biplot shows that the desirable genotype was the G4 (ARMA), followed by G2 (FACTOR), G3 (COSTANZA), G5 (PR31A34) and G1 (PR31Y43). All the aforementioned genotypes were positioned within the concentric area surrounding the optimum genotype.

5. Conclusions

According to our results and approach, protein content is considered a less qualitative characteristic than other characteristics like moisture and starch content. Therefore, negative selection for the last two characteristics and for ash content, combined with longer seeds, may indirectly lead to improved stability performance for protein content.

Hybrid Mitic showed higher values for almost all characteristics, indicating great adaptability and stability. Lines of the F-series also demonstrated satisfactory results, indicating a stability potential. Hybrid Kermes showed high stability for protein content.

The optimal genotypes to employ and environments/types of cultivation were determined by AMMI and GGE biplot. The genotypes G28 (8F), G27 (7F), G23 (3F), G8 (DKC6818), and G6 (ELEONORA) were relatively stable for moisture content, however, with lower performance for this characteristic. On the other hand, Genotypes G3 (COSTANZA) and G1 (PR31Y43) have the best seed size attributes for stability in the locations evaluated.